The RFQ Noise Floor

Priority poison

Years before I got into manufacturing I worked on advertising software, and I watched a routing queue die because of one override. The optimizer ran across campaign IDs, pacing curves, impression caps, and monthly budgets, and the single thing a human operator could touch was a priority flag we called BBB — Burn Baby, Burn. BBB shoved one campaign to the front of the delivery queue. The spend line climbed, the dashboard looked fixed, and everybody was happy for about a week. Then the other operators noticed and started flagging their own campaigns to protect them, and pretty soon the allocator could no longer tell priority from background noise. The inventory pool hadn't changed and neither had the model. The only thing that changed was that every campaign was now urgent.

We fixed it by killing BBB everywhere at once and forcing every campaign back through a single queue. It felt reckless at the time, because we were taking away the only lever anyone could see.

Manufacturing has the same disease and no single switch you can flip over a weekend. The working objects here are a drawing PDF, a STEP model, a revision letter, a tolerance block, a material spec, a finish note, a ship date, and an inspection plan — and priority doesn't live in a flag, it lives in behavior. A buyer blasts an RFQ to twenty shops, drops "urgent" in the subject line, and calls the owner's cell. The shop reshuffles the board and bumps a hot traveler ahead of work that was already released. Do that often enough and urgency stops carrying information at all, and the RFQ channel quits being a routing layer and turns into a noise source.

The failure both sides already recognize

Most RFQs are not demand. They are uncertainty being broadcast at scale.

The package shows up with a drawing PDF and a STEP file, and then the trouble starts: the revision letter is unstable, the note block fights the title block, the tolerance block is tight with no implied inspection method, and the cert requirement is either missing or living in somebody's tribal knowledge. The buyer isn't really asking for a price. He's asking a price to resolve an internal ambiguity he hasn't resolved himself, so the packet goes out to everyone who looks plausible on a capability list, because blasting is cheaper than pinning down datum intent or acceptance criteria.

Every wrong recipient still pays for it. The estimator opens the drawing, checks material and finish, scans the GD&T frames, and works out whether the part fits the machine envelope and the cert stack. Multiply that across the network and you've burned real estimator-hours before a single PO exists. That cost doesn't evaporate. It comes back as higher quote prices, longer quoted lead times, and eventually silence, once the inbox saturates.

Shops don't quote "parts." They process packets, and a packet has to be routable. Routable means an estimator can map the drawing to setups, fixturing, tool access, outside processes, and an inspection plan without guessing, because a guess becomes scrap, rework, or a missed ship date once the traveler is on the floor. Here's the constraint most buyers never see: in the U.S., quoting steals time from the same people who run a spindle, write a probe routine, or build a fixture. The quote queue is frequently tighter than the machine queue. A buyer stares at an unanswered email and concludes the part must be hard to make, when what actually happened is that the shop declined to do unpaid consulting on an ambiguous tolerance block and then chase him for answers. When the quote queue is the bottleneck, a shop protects it the way it protects any constrained cell — by filtering on completeness, buyer responsiveness, and expected award probability.

Why the U.S. inbox goes dark

Buyer expectations and supplier reality are on a collision course, and you can read the collision straight off the artifacts. Paperless Parts has published survey results showing that 67% of buyers expect a quote in under 24 hours, and only 6% are willing to wait longer than three days. That sets the default tempo for the entire channel. Now look at the other side of it: Paperless Parts marketing materials and a third party release cite up to two hours to quote a single component, and up to four days for complex assemblies — and that's before anyone has spent a minute on workholding, routing, or an inspection plan tied to a traveler. At twenty packets a day with a handful of "two hour" packets in the pile, the estimator queue becomes the hard limit, and silence is just what arithmetic looks like from the outside.

Overseas quoting behavior looks different because quote labor is cheaper and the quoting function is often staffed like a clerical pipeline. A supplier can fire back a number quickly because the cost of being wrong is deferred into assumptions, change orders, and loose inspection expectations, and what the buyer is really purchasing is a price signal rather than a stable work package. Most American shops can't run that play. The quote work competes directly with machine time and CMM programming, and the shop is the one holding a traveler with a rev letter, a cert packet, and a ship date on it. So U.S. suppliers get selective. They quote repeat buyers and ignore broadcast packets that smell like fishing. That's the black hole buyers complain about, and it is a defense mechanism.

The noise floor and the spiral

Every RFQ channel has a noise floor — the point where adding more RFQs produces fewer useful responses instead of more awards. Above that line, suppliers stop investing time in unclear drawings, response rates fall, and the best shops disengage first, because they already have stable work coming in through direct relationships with cleaner packets. Buyers react exactly the way you'd expect: widen the blast list, add more urgency, push more packets into more inboxes. Which increases triage labor. Which pushes more shops into silence.

It reads like a market problem, but the mechanism underneath is representation. A routing layer can't route what was never expressed as a complete packet, and it can't learn anything when silence disappears into email.

You can see the noise floor in the rates attached to an RFQ record and in the artifacts that follow it. A buyer with a high no-response rate isn't receiving supplier feedback; he's broadcasting packets that are too expensive to interpret relative to their expected award probability. A shop with low quote-to-PO conversion is burning estimator-hours on low quality demand, and it will eventually protect itself by quoting slower, quoting higher, or quoting only the part families it already knows. Once enough shops take that posture, the buyer experiences a market full of no-quotes while the machine capacity sits right there in the building.

The measurable signature

You don't need a moral diagnosis here. You need instrumentation on the RFQ record and the traveler. Track acknowledgement time from RFQ receipt to a first state change, because a healthy channel returns a quote, a question list tied to the drawing, or a decline with a reason tied to a constraint. Track the no-response rate, because silence is the cleanest early signal that the inbox has stopped behaving like a routing layer. Track quote-to-PO conversion, because low conversion means estimator-hours are going into uncertainty rather than demand, and you'll pay for that later in padded lead times and shops that won't engage. Add revision churn and PO amendments after traveler release, and you can see whether an unstable drawing packet is what's driving schedule thrash, fixture rework, and inspection re-planning.

Benchmarks vary by process and buyer mix, and many shops operate with bid-win rates far below 100% while a lot of quoting effort goes nowhere. One ERP vendor summarizes "industry studies" as bid-win rates around 30% with quote turnaround times in the three to four day range, which matches what estimators in custom fabrication describe as normal when the packet isn't clean. Honestly, though, you don't need anybody's benchmark. Your own RFQ records will tell you whether your no-response rate is climbing and whether your quote-to-PO conversion is collapsing.

Controls that touch the packet

Controls only work when they attach to something real: a drawing, an RFQ record, a PO, a traveler, an inspection plan. So require minimum packet fields before anything routes — drawing PDF with revision, tolerance block, material spec, finish spec, quantity, ship date, cert requirement, and a named buyer contact who can actually answer a question about datum intent and acceptance criteria. Require explicit supplier states on the RFQ record and keep the list short: quote, questions, decline, refer. Make a decline carry a reason tied to a constraint — machine envelope, process capability, cert stack, outside process availability, inspection burden. And close the award on the RFQ record even when no PO is issued, with a reason code: price, lead time, design change, program cancelled, internal make. Closure is what stops the same dead demand from generating unpaid engineering three more times.

Then stop letting expediting outrun revision control, which is how schedule boards and travelers lose their meaning in the first place. Lock the revision letter before you accept an expedite request. Make the question list part of the packet, so inspection planning gets tied to the tolerance block instead of improvised at the CMM. Cap hot travelers against setup capacity and inspection throughput, because priority always displaces something that was already planned, and a shop can only absorb so many disruptions before it starts producing partial work and missing inspections. Boring rules. They work because they're bolted to the artifacts that actually run the floor.

Why this matters

Domestic throughput gets destroyed before the first chip is cut, and it happens through a channel everyone treats as free. Estimator-hours disappear into drawings that are missing fields, unstable in revision, or unlikely to award — hours that would otherwise go into fixture design, probe routines, and inspection plans that raise first-pass yield on the traveler. The buyer sees the symptoms downstream: no-quotes, padded lead times, pricing that looks crazy. He reads all of it as a capacity shortage, when what he's looking at is a routing failure. The BBB story is useful because it strips the mechanism down to one variable. Mark every packet urgent and urgency stops being information, and the allocator collapses into noise.

Paper 7 moves from the RFQ inbox out to the supply chain that's already there and hard to see — the travelers, cert packets, and inspection reports moving between firms, and the identity and evidence you'd need to route work without broadcasting uncertainty.